The alpha risk of the Taguchi method for STB type quality characteristic

Abstract

One of the basic assumptions of experimental design is that the error variances are equal for all treatment combinations. On the contrary, one of the basic assumptions of Taguchi’s parameter design is that the error variances are not equal for the treatment combinations. Thus, the significant parameter levels are found by maximising the signal-to-noise ratio of the quality characteristic. In the analysis of variance (ANOVA) of the signal-to-noise ratio, the combination of column effects to better estimate error variance is referred to as pooling. Taguchi has suggested the strategy of “pooling-up”. When using the pooling-up strategy, there will be the tendency to make the alpha mistake more often. If the assumption of the former is true, then there is an alpha risk that judges some factor being significant when in fact it is not. The purpose of this paper is to investigate the alpha risk of the Taguchi method for the smaller-the-better (STB) type problem by simulation. The results show that the alpha risk is very high for several orthogonal arrays.